by: Rebecca B. Walzak, president and CEO of Walzak Risk Analysis
Concerned that increasing mortgage defaults will harm securities backed by those loans, investors have been increasingly demanding that lenders repurchase massive amounts of loans with early payment defaults, loans that were at least 90 days delinquent during the first year of performance. Subprime lenders unable to afford those buybacks have closed shop in droves, undertaken massive layoffs, and headed to the auction block. Media outlets, from trade publications to newspapers and national business magazines, are filled with stories of the so-called "subprime meltdown," with some even predicting a carnage that will drag the entire economy into recession.
According to the common belief, those defaults were caused by risky mortgage products - low-doc and no doc loans, high loan-to-values and interest-only products marketed to unsophisticated subprime borrowers. At least that's the widespread belief. But many of those bad loans are due to defects in lenders' operational processes - not defects in the loan products themselves. How can MBS investors be sure that a security is, for instance, backed by mortgages with a weighted average 80 percent loan-to-value, and homeowners with 650 FICO scores and verified income and employment? Just because the lender says the loans meet those characteristics does not make them so. The fact is: loan files are filled with mistakes. In some lending shops process errors may be pervasive.
The list of what can go wrong is long indeed. Loan files may suffer from misinformation, miscalculations and misapplications - intentional or not - in any point in process, from taking borrower data to servicing set up. For instance, the borrower may claim an income of $6,000 a month when that income was only for 10 months. The lender's staff may make typographical errors or mix up files. The loan officer may calculate income incorrectly. Or in a case of misapplying underwriting guidelines, the loan officer may accept two months of reserves when the guidelines require three.
As the refi boom gained steam in 2002 and 2003, lenders scrambled to meet rising demand and find bodies to push loans through the process. If you could add two numbers together you could be hired as a loan processor, and if you were a receptionist you could become an underwriter two weeks later. As they hired inexperienced personnel and pushed their systems to their limits, the likelihood for errors increased. The result is that securities backed by mortgages originated during the refinance boom are even more likely to have errors.
Under the common view, fraud shoulders much of the blame for defaults. Yet fraud is really due to faulty lending processes that should have caught fraudulent loans in the first place.
A clear relationship
The relationship between poor lender operational processes and loan performance is clear: defective processes mean higher defaults. In a recent study, Fitch Ratings has stated that about 30% of defaults can be due to mistakes in lender operational processes. And as defaults mount, investors reexamine files in hopes of finding mistakes that would allow them to force bad loans back to lenders.
Other industries have long understood operational risk. If you don't complete the steps correctly, the finished product, whatever that product may be, will not be what is expected. The government also recognizes its importance, employing the Sarbanes-Oxley Act to require companies to examine their operational processes. While other industries have embraced Six Sigma, lender process risk has so far gone largely unexamined.
The current quality control process as required by Department of Housing and Urban Development and Fannie Mae and Freddie Mac is inadequate for the task of measuring lender operational process risk. Manually rechecking portfolios loan by loan is costly and time consuming. By the time a file review is completed, the loan may already be sold - or may have already defaulted. Loan by loan reviews can identify and correct errors in individual loan files, but they cannot find defects in lenders' overall operational processes, shortcomings that produced those mistakes in the first place.
In addition, manual reviews are not objective or standardized. Reviewers bring all their human prejudices and preconceptions to the table. For instance, how important is an incorrect income? Or a missing employment verification? Being human, each reviewer places his or her own importance on these questions. They cannot determine the increased likelihood that missing or incorrect information will mean a default for a particular loan, not to mention forecast defaults for a portfolio. The lack of standardization in the current manual review process causes difficulty in the secondary market. Investors cannot easily compare the quality between different originators, and originators cannot readily compare themselves against competitors.
Lack of automation and standardization has so far slowed the advancement in lender process risk management the way lack of hard numbers and standardization held back credit reviews. Investors and lenders have always considered the character of borrowers and, with it, the probability that they will repay the loan, but the acceptance of FICO scores enabled credit checks to be automated and standardized, an advancement that boosted the lending and secondary market industries. A similar quantification of lender processes can bring similar benefits.
Others risks, such as interest rate, collateral and credit risk, are based on hard numbers. But lender operational risk has never been closely examined, not to mention quantified. These risks, like others, can be identified, controlled, monitored and priced.
Despite the pervasiveness of process defects, the industry has yet to determine their exact impact on loan performance and to compare it to the impact of poor loan products. How many defaults are due to process defects and how many unwise products? To answer that question, you must control for operational risk, identify mistakes and determine their level of default risk, not just loan by loan, but for the entire portfolio and the lender's entire process. Investors typically find an infraction of underwriting requirements when they demand a repurchase. Yet they usually reject loans without knowing if the error caused the default or understanding how those errors relate to the lender's overall performance. They don't know if a particular mistake was due to faulty lender procedures and if it is common in the lender's production.
Lenders and investors must determine the frequency of particular mistakes, not just identify and correct individual mistakes. For instance, suppose appraisal reviews report that 25% of appraisals show mistakes that question the value of the property. But the reviews find no common appraiser or underwriter, and the lender can easily obtain another comp and correct a valuation. However, the process lacks the necessary steps to effectively review appraisals. Although appraisals may be off only $200, but if the lender gets a $2 million loan, a mistake could mean a critical difference on both the probability of default and the amount of the loss.
Automation of operational risk will not make traditional quality control obsolete. Rather, it will allow QC to focus on lenders and loan files that need further investigation without wasting unnecessary scrutiny on high-quality portfolios, much like automated underwriting has allowed underwriters to spend more time on more unusual and problematic loans.
Building such an automated review is not an easy task. In order to verify data in the lender's files, the solution requires a partnership with specialists providing collateral, compliance, title, and fraud detection data, in addition to a slew of borrower data, such as Social Security numbers and income verification from numerous sources. Decisioning software is needed to automate the reviews, matching the data against outside sources and quantifying the risk.
Hard numbers on lender operational risk can help predict loan performance much like credit scores. Being automated, a number-based rating of operational process is streamlined and fast.
Investors could use the measurement to grade originators, avoiding ones with poor processes and embracing those with higher quality, or at least paying more to higher-quality originators and less for lower quality.
Lenders, for their part, could use automated reviews to avoid repurchases that have plagued today's market by improving their processes before the loans are sold on the secondary market. Despite assertions that lenders receive improved pricing for quality, the reality is that volume wins better pricing. But lenders could also use a standardized, third-party score to point out their superior quality to investors, and thus win superior pricing.
A score can help meet the regulatory requirement that lenders have control over their process and help discourage legislation for a suitability requirement. Guidance that
federal regulators issued on nontraditional products states that lenders must demonstrate they have control over their processes. Simply saying that you don't finance risky loans to unsophisticated consumers is not enough. Just listing underwriting criteria will not satisfy regulators. Lenders must show that they follow their guidelines. That kind of demonstration will help stave off a suitability standard brewing in Congress.
Although risky products may have been marketed to unsophisticated consumers, we don't know how many defaults are directly attributable to those bad products and how are due to operational defects. If we assume that all defaults are due to risky products without examining lenders' operational processes, we could eliminate or unnecessarily curtail loan products and thus cut off credit to many consumers.
Walzak Risk Analysis is a Boca Raton, Fla.-based company specializing in risk management of lenders' operational processes. CEO Walzak can be reached at 561-367-7333 or email@example.com.
(c) 2007 Asset Securitization Report and SourceMedia, Inc. All Rights Reserved.